29 March 2013 Fuzzy object models for newborn brain MR image segmentation
Author Affiliations +
Proceedings Volume 8672, Medical Imaging 2013: Biomedical Applications in Molecular, Structural, and Functional Imaging; 86720J (2013); doi: 10.1117/12.2001161
Event: SPIE Medical Imaging, 2013, Lake Buena Vista (Orlando Area), Florida, United States
Abstract
Newborn brain MR image segmentation is a challenging problem because of variety of size, shape and MR signal although it is the fundamental study for quantitative radiology in brain MR images. Because of the large difference between the adult brain and the newborn brain, it is difficult to directly apply the conventional methods for the newborn brain. Inspired by the original fuzzy object model introduced by Udupa et al. at SPIE Medical Imaging 2011, called fuzzy shape object model (FSOM) here, this paper introduces fuzzy intensity object model (FIOM), and proposes a new image segmentation method which combines the FSOM and FIOM into fuzzy connected (FC) image segmentation. The fuzzy object models are built from training datasets in which the cerebral parenchyma is delineated by experts. After registering FSOM with the evaluating image, the proposed method roughly recognizes the cerebral parenchyma region based on a prior knowledge of location, shape, and the MR signal given by the registered FSOM and FIOM. Then, FC image segmentation delineates the cerebral parenchyma using the fuzzy object models. The proposed method has been evaluated using 9 newborn brain MR images using the leave-one-out strategy. The revised age was between -1 and 2 months. Quantitative evaluation using false positive volume fraction (FPVF) and false negative volume fraction (FNVF) has been conducted. Using the evaluation data, a FPVF of 0.75% and FNVF of 3.75% were achieved. More data collection and testing are underway.
© (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Syoji Kobashi, Jayaram K. Udupa, "Fuzzy object models for newborn brain MR image segmentation", Proc. SPIE 8672, Medical Imaging 2013: Biomedical Applications in Molecular, Structural, and Functional Imaging, 86720J (29 March 2013); doi: 10.1117/12.2001161; https://doi.org/10.1117/12.2001161
PROCEEDINGS
6 PAGES


SHARE
KEYWORDS
Fuzzy logic

Image segmentation

Brain

Magnetic resonance imaging

Neuroimaging

Medical imaging

Object recognition

Back to Top